How to build and configure an agent in Flo AI?

Flo AI agents act as smart, dedicated teammates that help automate your procurement workflows, monitor requests, and keep your SaaS operations running smoothly. By configuring custom AI agents, you can streamline routine tasks and ensure your procurement processes align with organizational goals.

Where to find it

Agent configuration lives in Agent Studio, under the Flo AI section of the left-hand navigation — not under platform Settings.

Step 1. Open Agent Studio

Click Flo AI in the left nav, then Agent Studio. This lists every agent configured for your organization, with its status (e.g. Active).

Step 2. Start a new agent

Click + Create Agent in the top-right corner.

Step 3. Configure your agent

A new, unpublished agent (v1.0) opens on the Build tab:

• Name — starts as “Untitled Agent”; click to rename.
• + Add description — optional, helps others understand what the agent is for.
• Instructions (required) — plain-language directions for what the agent should do. You can reference platform data, variables, and uploaded documents inline using @ — see Writing agent instructions: @ mentions and formatting for the full picker reference.
• Domain (right panel) — the set of Spendflo entities (Suppliers, Contracts, Invoices, etc.) this agent is allowed to access. Only entities selected here are available to the agent.
• Variables (right panel) — placeholders you can reference in your instructions with @. If the agent is invoked from a workflow step, these get filled in automatically with that step's data.

Step 4. Save or publish

Click Save Only to keep working on it as a draft, or Publish to make it live and usable in workflows and Copilot right away. For how versioning, publishing, and re-editing a live agent works, see Versioning, publishing, and deleting custom agents.

Next steps

Now that you've configured your AI agent, explore:

• Writing agent instructions: @ mentions and formatting
• Reference: what data your agent can access
• Using the built-in @instance variable
• Versioning, publishing, and deleting custom agents
• Managing your Context Library
• Adding a Flo Agent step to a workflow
• Viewing and acting on a Flo Agent step on a request
• Audit trail for AI agent activity